What exactly do year to year changes in species index values stand for?
Such changes stand for fluctuations in population numbers, which we should interpret with some caution. PECBMS aims to detect long term rather than short term trends in population indices. Year to year changes, even when large, have limited relevancy to conservation. For instance, some bird species drop in numbers in reaction on extreme whether events (e.g. hard winters), but their populations may recover very quickly. Some species fluctuate in numbers naturally because of fluctuating food availability.
Added to such natural yearly fluctuations comes statistical noise. Therefore one should pay attention to the standard error of an index or trend. Values with large standard errors are less precise and indices based on such values will fluctuate more strongly.
What is TRIM, how does it work, where can it be obtained?
TRIM (TRends and Indices for Monitoring Data; Pannekoek & Van Strien, 2001) is the standard software tool we use to analyze time series of count data obtained from monitoring schemes and to produce estimates of yearly indices and trends. The analysis is based on loglinear Poisson regression (a form of generalized linear modeling). Such analyses are also available in standard statistical packages, but because TRIM is tailored to a selected set of models only, it runs faster and is easier to use.
TRIM is a freeware program, developed by Statistics Netherlands to be applied in wildlife statistics. It can be downloaded on http://www.ebcc.info/trim.html.
There is some evidence that climate change causes shifts in the phenology of species. Is it possible that a detected decline in numbers of a species is caused by the fact that the observers missed birds because these birds nowadays are singing or nesting a few weeks earlier than in former years?
No, this is highly improbable.
Breeding bird surveys are designed to span the full breeding season of birds from spring through to summer. They typically involve between two and 12 visits to a sample site per year through the breeding season to count birds. In this way, bird surveys span the full period when birds might be singing and nesting, and any shifts to early nesting would lie well within this period. Typically, the observed shifts to early nesting have been around 8 days in the UK so they are still relatively small in relation to the length of the breeding season and survey activity.
Note also that when a survey has just two site visits per year, one ´early´ in the breeding season and one ´late´, the higher count from the two visits is taken as the best measure of that species´ bird abundance on that square, and is used in subsequent analysis. This is the case for the Breeding Bird Survey used in the UK and several other European countries. The use of a maximum figure per species thus avoids the potential problem that earlier nesting might drag the mean count per visit downwards through time. We will however keep this methodological issue under review and check for bias.
Sampling the same sites every year is fine for farmland that is maintained, but in forests the species composition will change in the course of 25 years due to succession. Do sampling sites have to be replaced if they change strongly?
Sites are not replaced if they change, they continue to be monitored. This is because we aim to monitor changes in bird populations across large spatial and temporal scales as a consequence of changes in habitats, and forest succession is such a habitat change. To this end, we need a representative sample of sites across a wider area (e.g. a country) and if the forests in a country get older on average, specialists of older forest stages should increase. Such a phenomenon was shown for instance for forest birds in the Czech Republic (Reif et al., 2007).
There is a turnover of volunteer observers in every scheme. Some volunteers stop after some time and others enter the scheme. Doesn´t this turnover affect the results, given that volunteers have different observational skills?
No, it does not.
First, differences between observation skills are minimized because all field workers in each monitoring scheme adopt a standardized field method.
Second, potential differences among fieldworkers are addressed in the TRIM models used to compute indices (in statistical terms: the site-effect is taken into account in the model and site-effects include differences between observers).
Third, most differences between volunteers will add to the noise in the data and not to bias in the results. Only if observational skills are changing in a systematic way, bias can be expected. So far, there are hardly any examples of systematic changes in the PECBMS monitoring results that are to be ascribed to systematic changes in the skills of volunteers. It is however an ongoing concern of monitoring coordinators to avoid any systematic changes in observer skills.
Data are collected mainly by volunteers. How it is assured that volunteers properly collect the data? Do volunteer based monitoring schemes produce results that are as trustworthy as those from professional fieldworkers would be?
National organizers are keen to recruit volunteers with considerable experience in observing and recognizing birds. In fact, to produce reliable results, the involvement of volunteers is valuable, according to Schmeller et al. (2008):
“Because precision is a function of the number of monitored sites and the number of sites is maximized by volunteer involvement, our results do not support the common belief that volunteer-based schemes are too noisy to be informative. Just the opposite, we believe volunteer-based schemes provide relatively reliable data, with state-of-the-art survey designs or data-analysis methods, and consequently can yield unbiased results. Quality of data collected by volunteers is more likely determined by survey design, analytical methodology, and communication skills within the schemes rather than by volunteer involvement per se.”